Brownian Motion's Hidden Martingale Power

This question introduces the canonical continuous-time random process used in quantitative finance and stochastic modeling. The first part focuses on articulating what Brownian motion is and how its qualitative behavior is encoded in a handful of core properties. Candidates must explain how randomness unfolds over time in this model, and how its structure underlies diffusion-style dynamics used in derivatives pricing and risk modeling. The follow-up about relating the process to its own square pushes candidates to look beyond definitions and reason about functional transformations of stochastic processes.

Conceptually, the problem leans on understanding Gaussian increments, covariance structures, and how conditional expectations work for processes with independent increments. It implicitly uses basic martingale ideas and the connection between zero-drift, mean-zero increments and conditional expectations given the past. Strong answers identify which quantities are martingales, which are not, and why, and handle the dependence structure between a process and nonlinear functions of it. Interviewers watch for fluency with the defining toolkit of Brownian motion and the ability to turn those abstract properties into concrete probabilistic statements.

What it tests

The class of problems involving Brownian motion is governed by the interplay between independence of increments, Gaussian distributions, and time-homogeneity. The defining properties—stationary, independent increments and normality—mean that the entire process is determined by its covariance structure, which is linear in the minimum of times. This structure ensures that future evolution depends only on the present, not the path, embodying the Markov property. The continuous paths and scaling of variance with time are not arbitrary: they are consequences of the central limit theorem applied in continuous time, making Brownian motion the canonical model for random continuous evolution. The martingale property emerges because the process has no drift and its increments are mean-zero, so conditional expectations are always centered at the current value.